Among automated systems for detection of visual defects, techniques for car damage detection may serve a representative example.
Most of current damage assessment and reporting systems are based on the fact that the human eye is able to reliably detect and assess damage, if seen on a vehicle body, and record the results.
Some efforts have been made in the prior art to automate the process of car damage detection.
Various techniques of image comparison are available for the purpose of visual defects detection.
US2005125119 discloses a carriage that moves back and forth under a vehicle, and three cameras or a laser fixed along one axis pivot in a common plane, and teaches utilizing two cameras at time to locate a target fixed to a reference point on the undercarriage of the vehicle. Triangulation calculations, combined with the location of the cameras provide the location of the reference point in space in a three-dimensional coordinate system, compare that location with a stored designed location of the reference point prior to the crash, thus allowing the vehicle structure to be returned to its designed shape by other equipment.
US2006114531 describes methods, systems, and apparatuses for providing automated vehicle image acquisition, analysis, and reporting. One embodiment of the invention includes a method for providing vehicle damage information to an interested party. The method may include receiving simultaneously captured image data associated with a vehicle. The method can also include comparing at least a portion of the image data with previously stored image data associated with the vehicle. Furthermore, the method can include (based at least in part on the comparison of the portion of image data with previously stored image data associated with the vehicle), determining whether damage to the vehicle exists. In addition, the method can include outputting an indicator of whether damage to the vehicle exists.
US2007293997 describes a computer-assisted inspection system including an integrated software suite, which provides vehicle inspection services for various clients. The system provides customized workflow based on different inspection clients, different types of inspections and other factors. Different clients may have different inspection needs. Some clients may not care about certain defects that other clients want to know about. The system can provide, on the inspection appliance, different rules that can be activated in response to which client the inspection is being provided for and/or other factors for inspection criteria, valuation/pricing, rule-based validation and automatic grading. The resulting rule-driven user interface is thus customized-allowing the inspector to be more efficient in gathering the precise data desired at the time of data collection. Features include flat car part picking, highly client-customizability, easy to use, ability to run in an environment where the inspector has no constant connectivity to the network (no guaranteed access to the internet, full data replication, intermittent connectivity, synch back up), inspectors can be geographically separated (e.g., all over the country), system is installable over the internet to provide efficient installation to far-flung install sites.